Visual self-feedback method and device based on intelligent traction net rack

Abstract

The invention relates to the technical field of rehabilitation training, in particular to a visual self-feedback method and device based on an intelligent traction net rack, and the method comprises the steps that posture information of a patient is acquired, and the posture information comprises a target joint and a three-dimensional coordinate of the target joint; according to the three-dimensional coordinates, the real-time joint motion range of the target joint is calculated, and the joint motion range reflects the motion range and flexibility of the joint, namely the angle size and the motion amplitude which the joint can reach in different directions. According to the method, the posture of a patient is captured in real time through the depth camera, the joint motion range is accurately calculated, and personalized target traction force is dynamically generated based on a biomechanical model containing gravity compensation, elasticity and viscous resistance. Meanwhile, a dynamic safety threshold value changing along with the target force is created, accurate, safe and intelligent closed-loop control over the training process is achieved, and the rehabilitation effect and the patient safety are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation training technology. More specifically, this invention relates to a visual self-feedback method and device based on an intelligent traction frame. Background Technology

[0002] In the field of modern rehabilitation medicine, intelligent traction devices, such as suspension nets or rehabilitation robots, have become important tools to assist patients with lower limb dysfunction in rehabilitation training. These devices assist patients in completing specific training movements, such as hip flexion and extension, by applying precisely controlled traction force, aiming to restore joint range of motion and muscle strength.

[0003] However, current traction devices have limitations in force control strategies. Most systems use preset, constant traction force modes or only perform simple open-loop control, making it difficult to dynamically adjust the auxiliary force or resistance based on the patient's real-time posture and joint range of motion during training. This results in the training force failing to accurately compensate for the effects of changes in limb gravity, and lacks a safety monitoring mechanism that dynamically links to training goals. Inappropriate force may affect rehabilitation outcomes and even pose a risk of secondary injury. Summary of the Invention

[0004] This invention provides a visual self-feedback method and device based on an intelligent traction net frame, aiming to solve the problem that most systems in related technologies use preset, constant traction force modes or only perform simple open-loop control, making it difficult to dynamically adjust the auxiliary force or resistance according to the patient's real-time posture and joint range of motion during training.

[0005] In a first aspect, the present invention provides a visual self-feedback method and apparatus based on an intelligent traction frame, comprising: acquiring patient posture information, the posture information including a target joint and its three-dimensional coordinates; calculating the real-time joint range of motion of the target joint based on the three-dimensional coordinates, the joint range of motion reflecting the joint's range of motion and flexibility, i.e., the angle and amplitude that the joint can achieve in different directions; calculating a target traction force for training the patient based on the real-time joint range of motion, wherein the calculation of the target traction force involves a gravitational torque compensation term, an elastic resistance torque increment, and a viscous resistance term; generating a safety threshold based on the target traction force, and monitoring the actual traction force applied to the patient based on the safety threshold. By acquiring the patient's posture in real time and calculating the joint range of motion, the target traction force is decomposed into three terms—gravitational torque compensation, elastic resistance torque, and viscous resistance—for dynamic calculation, closed-loop adaptive control of the patient training process is achieved. This method can not only accurately compensate for the effects of changes in limb gravity and simulate more biomechanical assistance / resistance, but also generate a dynamic safety threshold based on dynamically calculated target traction force for real-time monitoring. This significantly improves the personalization, accuracy, and safety of rehabilitation training, and avoids poor rehabilitation results or the risk of secondary injury due to improper force control.

[0006] Furthermore, the real-time joint mobility of the target joint includes: constructing a first vector representing the direction of the patient's target limb segment and a second vector representing the direction of the patient's trunk or adjacent limb segments based on the three-dimensional coordinates of the target joint; calculating the angle between the first vector and the second vector, and using the angle as the joint mobility of the target joint. By constructing limb segment vectors and trunk vectors based on three-dimensional coordinates and calculating their angle to determine joint mobility, measurement errors caused by changes in camera viewpoint or slight patient positional movements can be eliminated, achieving more accurate and robust real-time quantification of joint mobility, and providing a reliable data foundation for subsequent precise force control calculations.

[0007] Furthermore, the gravitational torque compensation term is used to counteract the influence of the patient's lower limb weight. The magnitude of the gravitational torque compensation term is positively correlated with the patient's lower limb mass, the distance from the lower limb's center of mass to the target joint, and the cosine value of the joint's range of motion. By introducing a gravitational torque compensation term related to the cosine value of the real-time joint range of motion (angle), the torque effect generated by the patient's own weight at different lifting heights can be dynamically and accurately counteracted, ensuring that the applied traction force is purely used for the rehabilitation training goal, thereby improving the accuracy of the assistive force. Furthermore, the magnitude of the viscous resistance term is related to the preset viscous damping coefficient, the angular velocity of the target joint, and the joint range of motion.

[0008] Furthermore, the magnitude of the elastic resistance torque increment is positively correlated with the preset joint stiffness coefficient and the difference between the target value of the joint range of motion and the real-time joint range of motion.

[0009] Furthermore, it also includes adjusting or stopping the application of the target traction force if the monitored actual traction force exceeds the safety threshold, to ensure patient safety. By setting a closed-loop protection logic that automatically adjusts or stops the application of force when the monitored actual traction force exceeds the limit, a rapid and automatic response to abnormal situations is achieved. This greatly improves the safety of the rehabilitation training process and can effectively prevent harm to patients caused by unexpected situations such as sudden spasms or equipment malfunctions.

[0010] Furthermore, the safety threshold is calculated by multiplying the traction force of the motor by a preset safety factor to obtain the safety threshold.

[0011] Furthermore, the safety factor ranges from 1.5 to 2.0.

[0012] Furthermore, obtain the patient's posture information, including: Real-time acquisition of image data containing depth information using a depth camera; A pose estimation algorithm is used to identify and locate the target joint and its three-dimensional coordinates from the image data. In a second aspect, the present invention also provides a visual self-feedback method based on an intelligent traction net frame, comprising: a bed board 1, a support plate 2 installed on one side of the bed board 1, a first roller 3 and a second roller 4 installed on the support plate 2, a connecting rope provided on the first roller 3 and the second roller 4, a traction ring 7 installed at one end of the connecting rope, a sling 5 installed at the other end of the connecting rope, and two depth cameras 6 installed on the support plate 2 for collecting the patient's posture information.

[0013] Beneficial effects: By capturing the patient's posture in real time using a depth camera, the range of motion of the joints is accurately calculated, and a personalized target traction force is dynamically generated based on a biomechanical model that includes gravity compensation, elasticity, and viscous resistance. Simultaneously, a dynamic safety threshold that varies with the target force is created, achieving precise, safe, and intelligent closed-loop control of the training process, significantly improving rehabilitation outcomes and patient safety. Attached Figure Description

[0014] Figure 1 This is a schematic flowchart illustrating patient training according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the structure of a device according to an embodiment of the present invention.

[0015] Figure label: 1. Bed board; 2. Support board; 3. First roller; 4. Second roller; 5. Sling; 6. Depth camera; 7. Traction ring. Detailed Implementation

[0016] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0017] like Figure 1 As shown, S101: Collects the patient's posture information.

[0018] Specifically, at the start of rehabilitation training, the patient first lies on the bed surface of the device and the lower limb to be trained is fixed to the sling 5 by straps. Then, the depth camera or other equivalent RGB-D depth camera located above the net frame is activated to capture the patient's full-body posture in real time.

[0019] Understandably, depth cameras can simultaneously acquire color and depth images, thereby generating 3D point cloud data of the patient's body. The acquired image data stream can be processed using machine vision algorithms. As an exemplary implementation, a deep learning-based pose estimation algorithm (such as the OpenPose or MediaPipe framework) can be employed. This algorithm, trained on a large amount of human pose data, can accurately identify and locate the 3D spatial coordinates (X, Y, Z) of multiple key joints of the human body from images. In this embodiment, to achieve precise control of lower limb rehabilitation training, the system focuses on identifying joints related to lower limb movement, preferably including at least the hip, knee, and ankle joints. Simultaneously, to establish a body coordinate system benchmark, joints in the trunk, such as the shoulder joint, are also identified. The real-time 3D coordinates of these joints form the basis for all subsequent calculations.

[0020] S102: Calculate the patient's joint range of motion.

[0021] After obtaining the three-dimensional coordinates of the patient's key joints, it is necessary to calculate the patient's joint range of motion. Taking the thigh lifting (hip flexion) action as an example, the range of motion of this joint is calculated. Joint range of motion reflects the joint's range of motion and flexibility, that is, the angle and amplitude that the joint can achieve in different directions. Specifically, to quantify the patient's limb rehabilitation progress, this invention constructs vectors from the identified three-dimensional spatial joint coordinates and uses the vector dot product principle to accurately calculate the joint angle. This construction is based on the fact that the joint angle can be defined by the angle formed by the two body segments constituting the joint. For example, the hip flexion angle can be represented by the angle between the thigh and the torso.

[0022] For example, a formula for calculating the range of motion of the hip joint is provided: . The calculated real-time hip flexion angle; The vector pointing from the hip joint to the ankle joint represents the direction of the thigh (i.e., ankle joint coordinates - hip joint coordinates). The vector pointing from the hip joint to the shoulder joint represents the direction of the trunk (i.e., shoulder joint coordinates - hip joint coordinates). and These represent the lengths (magnitudes) of the vectors. In other words, the hip flexion angle is obtained by comparing the angle between two vectors pointing towards the thigh and the torso. When the patient raises their thigh, the position of the ankle joint changes, causing... The direction changes, thus causing the angle The changes in this angle allow for real-time tracking of the patient's hip joint range of motion.

[0023] S103: Calculate the target traction force to be applied to the patient at the next moment.

[0024] Specifically, taking the hip joint as an example, the range of motion and angular velocity of the hip joint are acquired in real time. Combined with the target angle difference for rehabilitation, the sum of the gravitational torque compensation term, the elastic resistance torque increment, and the viscous resistance term is calculated to obtain the total torque required for joint movement. Finally, this is converted into the traction force that the motor needs to apply. The conversion formula is as follows: In the formula, This indicates the target traction force output by the motor. This represents the total torque required for joint movement. This indicates mechanical efficiency, typically 0.9. This represents the vertical distance from the traction point to the hip joint. The vertical distance can be obtained by calculating it from real-time geometry (joint coordinates / linkage model) or by measuring it with a ruler when the posture is fixed.

[0025] A calculation formula is provided for the patient's gravitational torque compensation term. The formula is as follows: In the formula, For the patient's gravitational torque compensation term, For the patient's lower limb quality, It is the acceleration due to gravity. This is the distance from the center of mass of the lower limb to the hip joint. This refers to the range of motion of the hip joint. The patient's gravitational torque compensation term is calculated to counteract the effect of the lower limb's own weight on the hip joint. Under dynamic compensation, the traction force applied by the motor is more precise, ensuring that the assistive force is primarily used for rehabilitation training goals, rather than simply counteracting gravity.

[0026] A formula is provided for calculating the increment of elastic resistance torque for this patient: In the formula, For the patient's elastic resistance torque increment, The joint stiffness coefficient of the patient. This is the difference between the target value and the real-time joint range of motion. This indicates that the elastic response is adjusted based on the angle. The target value for joint range of motion is set manually based on experience.

[0027] A formula is provided for calculating the viscous resistance term for this patient: In the formula, For the patient's viscous resistance term, The viscous damping coefficient is... This refers to the range of motion of the hip joint. The value can be determined experimentally or based on empirical values. By controlling joint movement stability through speed- and angle-related resistance, the biomechanical realism of training is enhanced.

[0028] S104: Train the patient based on the target traction force.

[0029] Specifically, the calculated target traction force is sent as a command to the traction motor, which applies this force to the patient's lower limbs via ropes and straps to assist in training. Force sensors mounted on the traction ropes then monitor the actual force applied to the patient's limbs in real time. Finally, a safety threshold is dynamically calculated based on the sensor readings. The safety threshold is established based on the principle that under no circumstances should the applied force exceed a reasonable and safe range for the traction force. The calculation method for the safety threshold is as follows: In the formula, The safety threshold for applying resistance, The traction force applied to the patient, For safety reasons, in this embodiment, The value ranges from 1.5 to 2.0, preferably... The value is set to 1.7. This means that in actual training, the actual traction force is detected in real-time by the force sensor. It may exceed the traction force applied to the patient by the motor due to sudden spasms, unexpected resistance, mechanical jamming, or abnormal motor output. This indicates an abnormality or potential danger during training, requiring immediate protective measures. Without a safety threshold, it's impossible to determine when danger occurs; even if the actual output is significantly higher than the target force, the system may continue operating, threatening patient safety. Therefore, setting a safety threshold adds a dynamic monitoring standard to the training process, ensuring timely detection and protection of the patient even if the motor or patient's condition is abnormal.

[0030] In summary, by introducing a safety factor to construct a dynamic safety threshold, real-time monitoring and protection of traction force are achieved. This ensures absolute training safety while avoiding false alarms caused by slight force fluctuations, thus balancing the system's sensitivity and robustness. This threshold automatically adjusts with the target traction force, adapting to the needs of different patients and different rehabilitation stages, achieving personalized and adaptive control. This invention also provides a visual self-feedback device based on an intelligent traction grid structure. For example... Figure 2 As shown, the device includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement a visual self-feedback method based on an intelligent traction grid according to the first aspect of the present invention.

[0031] The visual self-feedback device based on the intelligent traction frame also includes: a bed board for the patient to lie on, with a protective pad installed on the side of the bed board that contacts the patient to improve comfort during training; a support plate installed on one side of the bed board to support it; and first and second rollers installed on the support plate, each with a connecting rope. One end of the connecting rope has a traction ring, and the other end has a sling for connecting to the patient's limbs. Two depth cameras are also installed on the support plate to collect the patient's posture information.

[0032] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0033] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A visual self-feedback method based on an intelligent traction grid structure, characterized in that, include: Acquire the patient's posture information, which includes the target joint and its three-dimensional coordinates; Based on the three-dimensional coordinates, the real-time joint mobility of the target joint is calculated. The joint mobility reflects the range of motion and flexibility of the joint, that is, the angle and range of motion that the joint can achieve in different directions. Based on the real-time joint range of motion, a target traction force for training the patient is calculated, wherein the calculation of the target traction force involves a gravitational torque compensation term, an elastic resistance torque increment, and a viscous resistance term; A safety threshold is generated based on the target traction force, and the actual traction force applied to the patient is monitored based on the safety threshold.

2. The visual self-feedback method based on an intelligent traction grid structure according to claim 1, characterized in that, The real-time range of motion of the target joint includes: Based on the three-dimensional coordinates of the target joint, a first vector representing the direction of the patient's target limb segment and a second vector representing the direction of the patient's trunk or adjacent limb segments are constructed. Calculate the angle between the first vector and the second vector, and use the angle as the joint range of motion of the target joint.

3. The visual self-feedback method based on an intelligent traction grid structure according to claim 1, characterized in that, The gravitational torque compensation term is used to counteract the influence of the patient's lower limb weight. The magnitude of the gravitational torque compensation term is positively correlated with the patient's lower limb mass, the distance from the lower limb's center of mass to the target joint, and the cosine value of the joint's range of motion.

4. The visual self-feedback method based on an intelligent traction grid structure according to claim 1, characterized in that, The magnitude of the viscous resistance term is related to the preset viscous damping coefficient, the angular velocity of the target joint, and the joint range of motion.

5. The visual self-feedback method based on an intelligent traction grid structure according to claim 1, characterized in that, The magnitude of the elastic resistance torque increment is positively correlated with the preset joint stiffness coefficient and the difference between the target value of the joint range of motion and the real-time joint range of motion.

6. The visual self-feedback method based on an intelligent traction grid structure according to claim 1, characterized in that, Also includes: If the actual traction force detected exceeds the safety threshold, the target traction force is adjusted or stopped to ensure patient safety.

7. The visual self-feedback method based on an intelligent traction grid structure according to claim 1, characterized in that, The security threshold is calculated as follows: The safety threshold is obtained by multiplying the traction force of the motor by a preset safety factor.

8. The visual self-feedback method based on an intelligent traction grid structure according to claim 7, characterized in that, The safety factor ranges from 1.5 to 2.

0.

9. The visual self-feedback method based on an intelligent traction grid structure according to claim 1, characterized in that, Obtain the patient's posture information, including: Real-time acquisition of image data containing depth information using a depth camera; A pose estimation algorithm is used to identify and locate the target joint and its three-dimensional coordinates from the image data.

10. A visual self-feedback device based on an intelligent traction grid structure, characterized in that, Also includes: A bed board (1) is provided, and a support plate (2) is also installed on one side of the bed board (1). A first roller (3) and a second roller (4) are also installed on the support plate (2). A connecting rope is provided on the first roller (3) and the second roller (4). A traction ring (7) is installed at one end of the connecting rope, and a sling (5) is installed at the other end of the connecting rope. Two depth cameras (6) are also provided on the support plate (2) for collecting the patient's posture information.

Images